mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Software (https://www.mersenneforum.org/forumdisplay.php?f=10)
-   -   Why it kept saying I got a PIII? (https://www.mersenneforum.org/showthread.php?t=36)

ebx 2002-08-26 21:20

Why it kept saying I got a PIII?
 
Hi,

Trying to run gimps at a new P4 2.2G box running FreeBSD 4.6. Every time I said its a P4, I came back to find it is a PIII.

My local.ini shows
CPUType=12
after I change but before long, it ia
CPUType=10
again.

Help!

Tasuke 2002-08-26 22:15

in version 22.X i noticed the cpu type will not change, but there is a new set of strings, OldCPUType and OldCPUSpeed. Those will have the correct values.

Prime95 2002-08-26 23:19

Try putting CpuSupportsSSE2=1 in local.ini. That *should* make it use the faster SSE2 code. Let me know if it does not.

Unfortunately, my only FreeBSD machine fried its disk a few months ago. Building a new v22 version will be a chore.

ebx 2002-08-27 01:08

Hmmm, it is worse

Timing 110 iterations at 256K FFT length. Illegal instruction (core dumped)

Here is part of my dmesg.

CPU: Pentium 4 (2220.56-MHz 686-class CPU)
Origin = "GenuineIntel" Id = 0xf24 Stepping = 4
Features=0x3febfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,DTS,ACPI,MMX,FXSR,SSE,SSE2,SS,<b28>,ACC>

Time to talk to the supplier?

Prime95 2002-08-27 01:19

I suspect FreeBSD is hosing you. The reason mprime was downgrading you to a P3 is that it tries to run an SSE2 instruction and it tries to core dump. mprime catches the signal and changes your cpu type from p4 to p3. The CpuSupportsSSE2=1 option tells mprime to use SSE2 instructions no matter what. And as we see it core dumps.

Is there any info on the FreeBsd web site regarding SSE2 enabled programs??

ebx 2002-08-27 03:10

Probably gcc too.

There is an option CPU_ENABLE_SSE I will enable it in the kernel.

ebx 2002-08-27 16:04

Yeah, CPU_ENABLE_SSE does the trick. At least now I dont need to switch to M$. :D

ebx 2002-08-30 21:04

The story hasnt ended. Sigh!

Benchmarking gives me expected numbers. It also uses this to estimate work so I got 5 exponents line up for double checking(default 10 day work). But when it comes to actual work, it is still like a P3, eg takes 5 days to do one test instead of 2 that I expect.

I even put CpuSupportsSSE2=1 to local.ini. Restarted the work. No difference. Do I need to wait until the current is over?

Prime95 2002-08-31 00:08

OK, first the obvious questions. No other tasks running stealing CPU time from mprime? Options/CPU reports SSE2 is supported? Does it report CMOV, SSE, PREFETCH are supported? Can you post your local.ini file?

Does Options/Benchmark report timings consistent with a P4 listed at http://www.mersenne.org/bench.htm?

You do not need to wait for the current test to complete to enjoy the faster iteration times. So something is still amiss.

ebx 2002-08-31 00:45

Lets see. I dont think there is anything even close to steal 3 day worth of CPU. Instead, it is almost idle other than my mprime. Here are the Options/CPU

Your choice: 15

CPU Type, 1=Pentium 4, 2=Pentium III, 3=Pentium II,
4=Celeron, 5=Pentium Pro, 6=Pentium, 7=486,
8=AMD Athlon, 9=AMD K6, 10=Cyrix (1):
CPU speed in MHz (2200):
Hours per day this program will run (24):

Please see the readme.txt file for very important
information on the available memory settings.

Daytime available memory in MB (300):
Nighttime available memory in MB (300):
Daytime begins at (7:30):
Daytime ends at (23:30):

Accept the answers above? (Y):

18. Options/Benchmark I didnt stop the background mprime.
Timing 110 iterations at 256K FFT length. Best time: 11.760 ms.
Timing 88 iterations at 320K FFT length. Best time: 15.041 ms.
Timing 73 iterations at 384K FFT length. Best time: 18.282 ms.
Timing 62 iterations at 448K FFT length. Best time: 21.855 ms.
Timing 55 iterations at 512K FFT length. Best time: 24.559 ms.
Timing 44 iterations at 640K FFT length. Best time: 31.732 ms.
Timing 36 iterations at 768K FFT length. Best time: 38.634 ms.
Timing 31 iterations at 892K FFT length. Best time: 47.148 ms.
Timing 27 iterations at 1024K FFT length. Best time: 51.309 ms.
Timing 22 iterations at 1280K FFT length. Best time: 72.146 ms.
Timing 18 iterations at 1536K FFT length. Best time: 88.063 ms.
Timing 15 iterations at 1792K FFT length. Best time: 107.193 ms.
Timing 110 iterations at 256K FFT length. Best time: 11.760 ms.
Timing 88 iterations at 320K FFT length. Best time: 15.041 ms.
Timing 73 iterations at 384K FFT length. Best time: 18.282 ms.
Timing 62 iterations at 448K FFT length. Best time: 21.855 ms.
Timing 55 iterations at 512K FFT length. Best time: 24.559 ms.
Timing 44 iterations at 640K FFT length. Best time: 31.732 ms.
Timing 36 iterations at 768K FFT length. Best time: 38.634 ms.
Timing 31 iterations at 892K FFT length. Best time: 47.148 ms.
Timing 27 iterations at 1024K FFT length. Best time: 51.309 ms.
Timing 22 iterations at 1280K FFT length. Best time: 72.146 ms.
Timing 18 iterations at 1536K FFT length. Best time: 88.063 ms.
Timing 15 iterations at 1792K FFT length. Best time: 107.193 ms.
Timing 110 iterations at 256K FFT length. Best time: 11.760 ms.
Timing 88 iterations at 320K FFT length. Best time: 15.041 ms.
Timing 73 iterations at 384K FFT length. Best time: 18.282 ms.
Timing 62 iterations at 448K FFT length. Best time: 21.855 ms.
Timing 55 iterations at 512K FFT length. Best time: 24.559 ms.
Timing 44 iterations at 640K FFT length. Best time: 31.732 ms.
Timing 36 iterations at 768K FFT length. Best time: 38.634 ms.
Timing 31 iterations at 892K FFT length. Best time: 47.148 ms.
Timing 27 iterations at 1024K FFT length. Best time: 51.309 ms.
Timing 22 iterations at 1280K FFT length. Best time: 72.146 ms.
Timing 18 iterations at 1536K FFT length. Best time: 88.063 ms.
Timing 15 iterations at 1792K FFT length. Best time: 107.193 ms.

I was getting 34ms for 256K table perviously.

ebx/mprime21> cat local.ini
CPUType=12
CPUSpeed=2200
CpuSupportsSSE2=1
RollingAverage=1061
RollingStartTime=1030753177
CPUHours=24
DayMemory=300
NightMemory=300
DayStartTime=450
DayEndTime=1410
ComputerID=ebx3
Pid=30314
LastEndDatesSent=1030742351
SelfTest448Passed=1
ebx/mprime21>

Snipped from boot up messages.

CPU: Pentium 4 (2220.57-MHz 686-class CPU)
Origin = "GenuineIntel" Id = 0xf24 Stepping = 4
Features=0x3febfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,DTS,ACPI,MMX,FXSR,SSE,SSE2,SS,<b28>,ACC>
real memory = 1065287680 (1040320K bytes)

I guess I will quit gimps after this current DC is done and rejoin. On my status, it says unknown CPU.

binarydigits 2002-08-31 01:13

[quote="ebx"]I guess I will quit gimps after this current DC is done and rejoin. On my status, it says unknown CPU.[/quote]

Don't worry about that, it isn't a problem on your end. Everyone's status page shows P4s as "unknown", it is a known bug on the Primenet server.


All times are UTC. The time now is 13:55.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2023, Jelsoft Enterprises Ltd.